<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<title>Lambert W function</title>
<link rel="stylesheet" href="../math.css" type="text/css">
<meta name="generator" content="DocBook XSL Stylesheets V1.79.1">
<link rel="home" href="../index.html" title="Math Toolkit 3.0.0">
<link rel="up" href="../special.html" title="Chapter 8. Special Functions">
<link rel="prev" href="jacobi_theta/jacobi_theta4.html" title="Jacobi Theta Function θ4">
<link rel="next" href="zetas.html" title="Zeta Functions">
</head>
<body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF">
<table cellpadding="2" width="100%"><tr>
<td valign="top"><img alt="Boost C++ Libraries" width="277" height="86" src="../../../../../boost.png"></td>
<td align="center"><a href="../../../../../index.html">Home</a></td>
<td align="center"><a href="../../../../../libs/libraries.htm">Libraries</a></td>
<td align="center"><a href="http://www.boost.org/users/people.html">People</a></td>
<td align="center"><a href="http://www.boost.org/users/faq.html">FAQ</a></td>
<td align="center"><a href="../../../../../more/index.htm">More</a></td>
</tr></table>
<hr>
<div class="spirit-nav">
<a accesskey="p" href="jacobi_theta/jacobi_theta4.html"><img src="../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../special.html"><img src="../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../index.html"><img src="../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="zetas.html"><img src="../../../../../doc/src/images/next.png" alt="Next"></a>
</div>
<div class="section">
<div class="titlepage"><div><div><h2 class="title" style="clear: both">
<a name="math_toolkit.lambert_w"></a><a class="link" href="lambert_w.html" title="Lambert W function">Lambert <span class="emphasis"><em>W</em></span>
function</a>
</h2></div></div></div>
<h5>
<a name="math_toolkit.lambert_w.h0"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.synopsis"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.synopsis">Synopsis</a>
</h5>
<pre class="programlisting"><span class="preprocessor">#include</span> <span class="special"><</span><span class="identifier">boost</span><span class="special">/</span><span class="identifier">math</span><span class="special">/</span><span class="identifier">special_functions</span><span class="special">/</span><span class="identifier">lambert_w</span><span class="special">.</span><span class="identifier">hpp</span><span class="special">></span>
</pre>
<pre class="programlisting"><span class="keyword">namespace</span> <span class="identifier">boost</span> <span class="special">{</span> <span class="keyword">namespace</span> <span class="identifier">math</span> <span class="special">{</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">);</span> <span class="comment">// W0 branch, default policy.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">);</span> <span class="comment">// W-1 branch, default policy.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_w0_prime</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">);</span> <span class="comment">// W0 branch 1st derivative.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_wm1_prime</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">);</span> <span class="comment">// W-1 branch 1st derivative.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">,</span> <span class="keyword">const</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">&);</span> <span class="comment">// W0 with policy.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">,</span> <span class="keyword">const</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">&);</span> <span class="comment">// W-1 with policy.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_w0_prime</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">,</span> <span class="keyword">const</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">&);</span> <span class="comment">// W0 derivative with policy.</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">lambert_wm1_prime</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">z</span><span class="special">,</span> <span class="keyword">const</span> <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a><span class="special">&);</span> <span class="comment">// W-1 derivative with policy.</span>
<span class="special">}</span> <span class="comment">// namespace boost</span>
<span class="special">}</span> <span class="comment">// namespace math</span>
</pre>
<h5>
<a name="math_toolkit.lambert_w.h1"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.description"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.description">Description</a>
</h5>
<p>
The <a href="http://en.wikipedia.org/wiki/Lambert_W_function" target="_top">Lambert W
function</a> is the solution of the equation <span class="emphasis"><em>W</em></span>(<span class="emphasis"><em>z</em></span>)<span class="emphasis"><em>e</em></span><sup><span class="emphasis"><em>W</em></span>(<span class="emphasis"><em>z</em></span>)</sup> =
<span class="emphasis"><em>z</em></span>. It is also called the Omega function, the inverse of
<span class="emphasis"><em>f</em></span>(<span class="emphasis"><em>W</em></span>) = <span class="emphasis"><em>We</em></span><sup><span class="emphasis"><em>W</em></span></sup>.
</p>
<p>
On the interval [0, ∞), there is just one real solution. On the interval (-<span class="emphasis"><em>e</em></span><sup>-1</sup>,
0), there are two real solutions, generating two branches which we will denote
by <span class="emphasis"><em>W</em></span><sub>0</sub> and <span class="emphasis"><em>W</em></span><sub>-1</sub>. In Boost.Math, we call
these principal branches <code class="computeroutput"><span class="identifier">lambert_w0</span></code>
and <code class="computeroutput"><span class="identifier">lambert_wm1</span></code>; their derivatives
are labelled <code class="computeroutput"><span class="identifier">lambert_w0_prime</span></code>
and <code class="computeroutput"><span class="identifier">lambert_wm1_prime</span></code>.
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_w_graph.svg" align="middle"></span>
</p></blockquote></div>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_w_graph_big_w.svg" align="middle"></span>
</p></blockquote></div>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_w0_prime_graph.svg" align="middle"></span>
</p></blockquote></div>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_wm1_prime_graph.svg" align="middle"></span>
</p></blockquote></div>
<p>
There is a singularity where the branches meet at <span class="emphasis"><em>e</em></span><sup>-1</sup> ≅ <code class="literal">-0.367879</code>.
Approaching this point, the condition number of function evaluation tends to
infinity, and the only method of recovering high accuracy is use of higher
precision.
</p>
<p>
This implementation computes the two real branches <span class="emphasis"><em>W</em></span><sub>0</sub> and
<span class="emphasis"><em>W</em></span><sub>-1</sub>
with the functions <code class="computeroutput"><span class="identifier">lambert_w0</span></code>
and <code class="computeroutput"><span class="identifier">lambert_wm1</span></code>, and their
derivatives, <code class="computeroutput"><span class="identifier">lambert_w0_prime</span></code>
and <code class="computeroutput"><span class="identifier">lambert_wm1_prime</span></code>. Complex
arguments are not supported.
</p>
<p>
The final <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policy</a> argument is optional and can
be used to control how the function deals with errors. Refer to <a class="link" href="../policy.html" title="Chapter 21. Policies: Controlling Precision, Error Handling etc">Policies</a>
for more details and see examples below.
</p>
<h6>
<a name="math_toolkit.lambert_w.h2"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.applications"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.applications">Applications
of the Lambert <span class="emphasis"><em>W</em></span> function</a>
</h6>
<p>
The Lambert <span class="emphasis"><em>W</em></span> function has a myriad of applications.
<a href="http://www.apmaths.uwo.ca/~djeffrey/Offprints/W-adv-cm.pdf" target="_top">Corless
et al.</a> provide a summary of applications, from the mathematical, like
iterated exponentiation and asymptotic roots of trinomials, to the real-world,
such as the range of a jet plane, enzyme kinetics, water movement in soil,
epidemics, and diode current (an example replicated <a href="../../../example/lambert_w_diode.cpp" target="_top">here</a>).
Since the publication of their landmark paper, there have been many more applications,
and also many new implementations of the function, upon which this implementation
builds.
</p>
<h5>
<a name="math_toolkit.lambert_w.h3"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.examples"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.examples">Examples</a>
</h5>
<p>
The most basic usage of the Lambert-<span class="emphasis"><em>W</em></span> function is demonstrated
below:
</p>
<pre class="programlisting"><span class="preprocessor">#include</span> <span class="special"><</span><span class="identifier">boost</span><span class="special">/</span><span class="identifier">math</span><span class="special">/</span><span class="identifier">special_functions</span><span class="special">/</span><span class="identifier">lambert_w</span><span class="special">.</span><span class="identifier">hpp</span><span class="special">></span> <span class="comment">// For lambert_w function.</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">lambert_w0</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">lambert_wm1</span><span class="special">;</span>
</pre>
<pre class="programlisting"><span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
<span class="comment">// Show all potentially significant decimal digits,</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">showpoint</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// and show significant trailing zeros too.</span>
<span class="keyword">double</span> <span class="identifier">z</span> <span class="special">=</span> <span class="number">10.</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span> <span class="comment">// Default policy for double.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(z) = "</span> <span class="special"><<</span> <span class="identifier">r</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(z) = 1.7455280027406994</span>
</pre>
<p>
Other floating-point types can be used too, here <code class="computeroutput"><span class="keyword">float</span></code>,
including user-defined types like <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>.
It is convenient to use a function like <code class="computeroutput"><span class="identifier">show_value</span></code>
to display all (and only) potentially significant decimal digits, including
any significant trailing zeros, (<code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">max_digits10</span></code>) for the type <code class="computeroutput"><span class="identifier">T</span></code>.
</p>
<pre class="programlisting"><span class="keyword">float</span> <span class="identifier">z</span> <span class="special">=</span> <span class="number">10.F</span><span class="special">;</span>
<span class="keyword">float</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span> <span class="comment">// Default policy digits10 = 7, digits2 = 24</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0("</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">") = "</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">r</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// lambert_w0(10.0000000) = 1.74552798</span>
</pre>
<p>
Example of an integer argument to <code class="computeroutput"><span class="identifier">lambert_w0</span></code>,
showing that an <code class="computeroutput"><span class="keyword">int</span></code> literal is
correctly promoted to a <code class="computeroutput"><span class="keyword">double</span></code>.
</p>
<pre class="programlisting"><span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
<span class="keyword">double</span> <span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="number">10</span><span class="special">);</span> <span class="comment">// Pass an int argument "10" that should be promoted to double argument.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(10) = "</span> <span class="special"><<</span> <span class="identifier">r</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// lambert_w0(10) = 1.7455280027406994</span>
<span class="keyword">double</span> <span class="identifier">rp</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="number">10</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(10) = "</span> <span class="special"><<</span> <span class="identifier">rp</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(10) = 1.7455280027406994</span>
<span class="keyword">auto</span> <span class="identifier">rr</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="number">10</span><span class="special">);</span> <span class="comment">// C++11 needed.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(10) = "</span> <span class="special"><<</span> <span class="identifier">rr</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(10) = 1.7455280027406994 too, showing that rr has been promoted to double.</span>
</pre>
<p>
Using <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
types to get much higher precision is painless.
</p>
<pre class="programlisting"><span class="identifier">cpp_dec_float_50</span> <span class="identifier">z</span><span class="special">(</span><span class="string">"10"</span><span class="special">);</span>
<span class="comment">// Note construction using a decimal digit string "10",</span>
<span class="comment">// NOT a floating-point double literal 10.</span>
<span class="identifier">cpp_dec_float_50</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0("</span><span class="special">;</span> <span class="identifier">show_value</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">") = "</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">r</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(10.000000000000000000000000000000000000000000000000000000000000000000000000000000) =</span>
<span class="comment">// 1.7455280027406993830743012648753899115352881290809413313533156980404446940000000</span>
</pre>
<div class="warning"><table border="0" summary="Warning">
<tr>
<td rowspan="2" align="center" valign="top" width="25"><img alt="[Warning]" src="../../../../../doc/src/images/warning.png"></td>
<th align="left">Warning</th>
</tr>
<tr><td align="left" valign="top"><p>
When using multiprecision, take very great care not to construct or assign
non-integers from <code class="computeroutput"><span class="keyword">double</span></code>, <code class="computeroutput"><span class="keyword">float</span></code> ... silently losing precision. Use
<code class="computeroutput"><span class="string">"1.2345678901234567890123456789"</span></code>
rather than <code class="computeroutput"><span class="number">1.2345678901234567890123456789</span></code>.
</p></td></tr>
</table></div>
<p>
Using multiprecision types, it is all too easy to get multiprecision precision
wrong!
</p>
<pre class="programlisting"><span class="identifier">cpp_dec_float_50</span> <span class="identifier">z</span><span class="special">(</span><span class="number">0.7777777777777777777777777777777777777777777777777777777777777777777777777</span><span class="special">);</span>
<span class="comment">// Compiler evaluates the nearest double-precision binary representation,</span>
<span class="comment">// from the max_digits10 of the floating_point literal double 0.7777777777777777777777777777...,</span>
<span class="comment">// so any extra digits in the multiprecision type</span>
<span class="comment">// beyond max_digits10 (usually 17) are random and meaningless.</span>
<span class="identifier">cpp_dec_float_50</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0("</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">") = "</span><span class="special">;</span> <span class="identifier">show_value</span><span class="special">(</span><span class="identifier">r</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(0.77777777777777779011358916250173933804035186767578125000000000000000000000000000)</span>
<span class="comment">// = 0.48086152073210493501934682309060873341910109230469724725005039758139532631901386</span>
</pre>
<div class="note"><table border="0" summary="Note">
<tr>
<td rowspan="2" align="center" valign="top" width="25"><img alt="[Note]" src="../../../../../doc/src/images/note.png"></td>
<th align="left">Note</th>
</tr>
<tr><td align="left" valign="top"><p>
See spurious non-seven decimal digits appearing after digit #17 in the argument
0.7777777777777777...!
</p></td></tr>
</table></div>
<p>
And similarly constructing from a literal <code class="computeroutput"><span class="keyword">double</span>
<span class="number">0.9</span></code>, with more random digits after digit
number 17.
</p>
<pre class="programlisting"><span class="identifier">cpp_dec_float_50</span> <span class="identifier">z</span><span class="special">(</span><span class="number">0.9</span><span class="special">);</span> <span class="comment">// Construct from floating_point literal double 0.9.</span>
<span class="identifier">cpp_dec_float_50</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="number">0.9</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0("</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">") = "</span><span class="special">;</span> <span class="identifier">show_value</span><span class="special">(</span><span class="identifier">r</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// lambert_w0(0.90000000000000002220446049250313080847263336181640625000000000000000000000000000)</span>
<span class="comment">// = 0.52983296563343440510607251781038939952850341796875000000000000000000000000000000</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(0.9) = "</span> <span class="special"><<</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="number">0.9</span><span class="special">))</span>
<span class="comment">// lambert_w0(0.9)</span>
<span class="comment">// = 0.52983296563343441</span>
<span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Note how the <code class="computeroutput"><span class="identifier">cpp_float_dec_50</span></code>
result is only as correct as from a <code class="computeroutput"><span class="keyword">double</span>
<span class="special">=</span> <span class="number">0.9</span></code>.
</p>
<p>
Now see the correct result for all 50 decimal digits constructing from a decimal
digit string "0.9":
</p>
<pre class="programlisting"><span class="identifier">cpp_dec_float_50</span> <span class="identifier">z</span><span class="special">(</span><span class="string">"0.9"</span><span class="special">);</span> <span class="comment">// Construct from decimal digit string.</span>
<span class="identifier">cpp_dec_float_50</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0("</span><span class="special">;</span>
<span class="identifier">show_value</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">") = "</span><span class="special">;</span> <span class="identifier">show_value</span><span class="special">(</span><span class="identifier">r</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// 0.90000000000000000000000000000000000000000000000000000000000000000000000000000000)</span>
<span class="comment">// = 0.52983296563343441213336643954546304857788132269804249284012528304239956413801252</span>
</pre>
<p>
Note the expected zeros for all places up to 50 - and the correct Lambert
<span class="emphasis"><em>W</em></span> result!
</p>
<p>
(It is just as easy to compute even higher precisions, at least to thousands
of decimal digits, but not shown here for brevity. See <a href="../../../example/lambert_w_simple_examples.cpp" target="_top">lambert_w_simple_examples.cpp</a>
for comparison of an evaluation at 1000 decimal digit precision with <a href="http://www.wolframalpha.com/" target="_top">Wolfram Alpha</a>).
</p>
<p>
Policies can be used to control what action to take on errors:
</p>
<pre class="programlisting"><span class="comment">// Define an error handling policy:</span>
<span class="keyword">typedef</span> <span class="identifier">policy</span><span class="special"><</span>
<span class="identifier">domain_error</span><span class="special"><</span><span class="identifier">throw_on_error</span><span class="special">>,</span>
<span class="identifier">overflow_error</span><span class="special"><</span><span class="identifier">ignore_error</span><span class="special">></span> <span class="comment">// possibly unwise?</span>
<span class="special">></span> <span class="identifier">my_throw_policy</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
<span class="comment">// Show all potentially significant decimal digits,</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">showpoint</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// and show significant trailing zeros too.</span>
<span class="keyword">double</span> <span class="identifier">z</span> <span class="special">=</span> <span class="special">+</span><span class="number">1</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// Lambert W (1.0000000000000000) = 0.56714329040978384</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"\nLambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">", my_throw_policy()) = "</span>
<span class="special"><<</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">,</span> <span class="identifier">my_throw_policy</span><span class="special">())</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// Lambert W (1.0000000000000000, my_throw_policy()) = 0.56714329040978384</span>
</pre>
<p>
An example error message:
</p>
<pre class="programlisting"><span class="identifier">Error</span> <span class="identifier">in</span> <span class="identifier">function</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">lambert_wm1</span><span class="special"><</span><span class="identifier">RealType</span><span class="special">>(<</span><span class="identifier">RealType</span><span class="special">>):</span>
<span class="identifier">Argument</span> <span class="identifier">z</span> <span class="special">=</span> <span class="number">1</span> <span class="identifier">is</span> <span class="identifier">out</span> <span class="identifier">of</span> <span class="identifier">range</span> <span class="special">(</span><span class="identifier">z</span> <span class="special"><=</span> <span class="number">0</span><span class="special">)</span> <span class="keyword">for</span> <span class="identifier">Lambert</span> <span class="identifier">W</span><span class="special">-</span><span class="number">1</span> <span class="identifier">branch</span><span class="special">!</span> <span class="special">(</span><span class="identifier">Try</span> <span class="identifier">Lambert</span> <span class="identifier">W0</span> <span class="identifier">branch</span><span class="special">?)</span>
</pre>
<p>
Showing an error reported if a value is passed to <code class="computeroutput"><span class="identifier">lambert_w0</span></code>
that is out of range, (and was probably meant to be passed to <code class="computeroutput"><span class="identifier">lambert_wm1</span></code> instead).
</p>
<pre class="programlisting"><span class="keyword">double</span> <span class="identifier">z</span> <span class="special">=</span> <span class="special">+</span><span class="number">1.</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_wm1(+1.) = "</span> <span class="special"><<</span> <span class="identifier">r</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
The full source of these examples is at <a href="../../../example/lambert_w_simple_examples.cpp" target="_top">lambert_w_simple_examples.cpp</a>
</p>
<h6>
<a name="math_toolkit.lambert_w.h4"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.diode_resistance"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.diode_resistance">Diode
Resistance Example</a>
</h6>
<p>
A typical example of a practical application is estimating the current flow
through a diode with series resistance from a paper by Banwell and Jayakumar.
</p>
<p>
Having the Lambert <span class="emphasis"><em>W</em></span> function available makes it simple
to reproduce the plot in their paper (Fig 2) comparing estimates using with
Lambert <span class="emphasis"><em>W</em></span> function and some actual measurements. The colored
curves show the effect of various series resistance on the current compared
to an extrapolated line in grey with no internal (or external) resistance.
</p>
<p>
Two formulae relating the diode current and effect of series resistance can
be combined, but yield an otherwise intractable equation relating the current
versus voltage with a varying series resistance. This was reformulated as a
generalized equation in terms of the Lambert W function:
</p>
<p>
Banwell and Jakaumar equation 5
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="serif_italic">I(V) = μ V<sub>T</sub>/ R <sub>S</sub> ․ W<sub>0</sub>(I<sub>0</sub> R<sub>S</sub> / (μ V<sub>T</sub>))</span>
</p></blockquote></div>
<p>
Using these variables
</p>
<pre class="programlisting"><span class="keyword">double</span> <span class="identifier">nu</span> <span class="special">=</span> <span class="number">1.0</span><span class="special">;</span> <span class="comment">// Assumed ideal.</span>
<span class="keyword">double</span> <span class="identifier">vt</span> <span class="special">=</span> <span class="identifier">v_thermal</span><span class="special">(</span><span class="number">25</span><span class="special">);</span> <span class="comment">// v thermal, Shockley equation, expect about 25 mV at room temperature.</span>
<span class="keyword">double</span> <span class="identifier">boltzmann_k</span> <span class="special">=</span> <span class="number">1.38e-23</span><span class="special">;</span> <span class="comment">// joules/kelvin</span>
<span class="keyword">double</span> <span class="identifier">temp</span> <span class="special">=</span> <span class="number">273</span> <span class="special">+</span> <span class="number">25</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">charge_q</span> <span class="special">=</span> <span class="number">1.6e-19</span><span class="special">;</span> <span class="comment">// column</span>
<span class="identifier">vt</span> <span class="special">=</span> <span class="identifier">boltzmann_k</span> <span class="special">*</span> <span class="identifier">temp</span> <span class="special">/</span> <span class="identifier">charge_q</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"V thermal "</span> <span class="special"><<</span> <span class="identifier">vt</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// V thermal 0.0257025 = 25 mV</span>
<span class="keyword">double</span> <span class="identifier">rsat</span> <span class="special">=</span> <span class="number">0.</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">isat</span> <span class="special">=</span> <span class="number">25.e-15</span><span class="special">;</span> <span class="comment">// 25 fA;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Isat = "</span> <span class="special"><<</span> <span class="identifier">isat</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">re</span> <span class="special">=</span> <span class="number">0.3</span><span class="special">;</span> <span class="comment">// Estimated from slope of straight section of graph (equation 6).</span>
<span class="keyword">double</span> <span class="identifier">v</span> <span class="special">=</span> <span class="number">0.9</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">icalc</span> <span class="special">=</span> <span class="identifier">iv</span><span class="special">(</span><span class="identifier">v</span><span class="special">,</span> <span class="identifier">vt</span><span class="special">,</span> <span class="number">249.</span><span class="special">,</span> <span class="identifier">re</span><span class="special">,</span> <span class="identifier">isat</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"voltage = "</span> <span class="special"><<</span> <span class="identifier">v</span> <span class="special"><<</span> <span class="string">", current = "</span> <span class="special"><<</span> <span class="identifier">icalc</span> <span class="special"><<</span> <span class="string">", "</span> <span class="special"><<</span> <span class="identifier">log</span><span class="special">(</span><span class="identifier">icalc</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// voltage = 0.9, current = 0.00108485, -6.82631</span>
</pre>
<p>
the formulas can be rendered in C++
</p>
<pre class="programlisting"><span class="keyword">double</span> <span class="identifier">iv</span><span class="special">(</span><span class="keyword">double</span> <span class="identifier">v</span><span class="special">,</span> <span class="keyword">double</span> <span class="identifier">vt</span><span class="special">,</span> <span class="keyword">double</span> <span class="identifier">rsat</span><span class="special">,</span> <span class="keyword">double</span> <span class="identifier">re</span><span class="special">,</span> <span class="keyword">double</span> <span class="identifier">isat</span><span class="special">,</span> <span class="keyword">double</span> <span class="identifier">nu</span> <span class="special">=</span> <span class="number">1.</span><span class="special">)</span>
<span class="special">{</span>
<span class="comment">// V thermal 0.0257025 = 25 mV</span>
<span class="comment">// was double i = (nu * vt/r) * lambert_w((i0 * r) / (nu * vt)); equ 5.</span>
<span class="identifier">rsat</span> <span class="special">=</span> <span class="identifier">rsat</span> <span class="special">+</span> <span class="identifier">re</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">i</span> <span class="special">=</span> <span class="identifier">nu</span> <span class="special">*</span> <span class="identifier">vt</span> <span class="special">/</span> <span class="identifier">rsat</span><span class="special">;</span>
<span class="comment">// std::cout << "nu * vt / rsat = " << i << std::endl; // 0.000103223</span>
<span class="keyword">double</span> <span class="identifier">x</span> <span class="special">=</span> <span class="identifier">isat</span> <span class="special">*</span> <span class="identifier">rsat</span> <span class="special">/</span> <span class="special">(</span><span class="identifier">nu</span> <span class="special">*</span> <span class="identifier">vt</span><span class="special">);</span>
<span class="comment">// std::cout << "isat * rsat / (nu * vt) = " << x << std::endl;</span>
<span class="keyword">double</span> <span class="identifier">eterm</span> <span class="special">=</span> <span class="special">(</span><span class="identifier">v</span> <span class="special">+</span> <span class="identifier">isat</span> <span class="special">*</span> <span class="identifier">rsat</span><span class="special">)</span> <span class="special">/</span> <span class="special">(</span><span class="identifier">nu</span> <span class="special">*</span> <span class="identifier">vt</span><span class="special">);</span>
<span class="comment">// std::cout << "(v + isat * rsat) / (nu * vt) = " << eterm << std::endl;</span>
<span class="keyword">double</span> <span class="identifier">e</span> <span class="special">=</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">eterm</span><span class="special">);</span>
<span class="comment">// std::cout << "exp(eterm) = " << e << std::endl;</span>
<span class="keyword">double</span> <span class="identifier">w0</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">x</span> <span class="special">*</span> <span class="identifier">e</span><span class="special">);</span>
<span class="comment">// std::cout << "w0 = " << w0 << std::endl;</span>
<span class="keyword">return</span> <span class="identifier">i</span> <span class="special">*</span> <span class="identifier">w0</span> <span class="special">-</span> <span class="identifier">isat</span><span class="special">;</span>
<span class="special">}</span> <span class="comment">// double iv</span>
</pre>
<p>
to reproduce their Fig 2:
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/diode_iv_plot.svg" align="middle"></span>
</p></blockquote></div>
<p>
The plotted points for no external series resistance (derived from their published
plot as the raw data are not publicly available) are used to extrapolate back
to estimate the intrinsic emitter resistance as 0.3 ohm. The effect of external
series resistance is visible when the colored lines start to curve away from
the straight line as voltage increases.
</p>
<p>
See <a href="../../../example/lambert_w_diode.cpp" target="_top">lambert_w_diode.cpp</a>
and <a href="../../../example/lambert_w_diode_graph.cpp" target="_top">lambert_w_diode_graph.cpp</a>
for details of the calculation.
</p>
<h6>
<a name="math_toolkit.lambert_w.h5"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.implementations"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.implementations">Existing
implementations</a>
</h6>
<p>
The principal value of the Lambert <span class="emphasis"><em>W</em></span> function is implemented
in the <a href="http://mathworld.wolfram.com/LambertW-Function.html" target="_top">Wolfram
Language</a> as <code class="computeroutput"><span class="identifier">ProductLog</span><span class="special">[</span><span class="identifier">k</span><span class="special">,</span>
<span class="identifier">z</span><span class="special">]</span></code>,
where <code class="computeroutput"><span class="identifier">k</span></code> is the branch.
</p>
<p>
The symbolic algebra program <a href="https://www.maplesoft.com" target="_top">Maple</a>
also computes Lambert <span class="emphasis"><em>W</em></span> to an arbitrary precision.
</p>
<h5>
<a name="math_toolkit.lambert_w.h6"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.precision"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.precision">Controlling
the compromise between Precision and Speed</a>
</h5>
<h6>
<a name="math_toolkit.lambert_w.h7"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.small_floats"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.small_floats">Floating-point
types <code class="computeroutput"><span class="keyword">double</span></code> and <code class="computeroutput"><span class="keyword">float</span></code></a>
</h6>
<p>
This implementation provides good precision and excellent speed for __fundamental
<code class="computeroutput"><span class="keyword">float</span></code> and <code class="computeroutput"><span class="keyword">double</span></code>.
</p>
<p>
All the functions usually return values within a few <a href="http://en.wikipedia.org/wiki/Unit_in_the_last_place" target="_top">Unit
in the last place (ULP)</a> for the floating-point type, except for very
small arguments very near zero, and for arguments very close to the singularity
at the branch point.
</p>
<p>
By default, this implementation provides the best possible speed. Very slightly
average higher precision and less bias might be obtained by adding a <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.halley">Halley</a> step refinement, but
at the cost of more than doubling the runtime.
</p>
<h6>
<a name="math_toolkit.lambert_w.h8"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.big_floats"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.big_floats">Floating-point
types larger than double</a>
</h6>
<p>
For floating-point types with precision greater than <code class="computeroutput"><span class="keyword">double</span></code>
and <code class="computeroutput"><span class="keyword">float</span></code> <a href="http://en.cppreference.com/w/cpp/language/types" target="_top">fundamental
(built-in) types</a>, a <code class="computeroutput"><span class="keyword">double</span></code>
evaluation is used as a first approximation followed by Halley refinement,
using a single step where it can be predicted that this will be sufficient,
and only using <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.halley">Halley</a>
iteration when necessary. Higher precision types are always going to be <span class="bold"><strong>very, very much slower</strong></span>.
</p>
<p>
The 'best' evaluation (the nearest <a href="http://en.wikipedia.org/wiki/Floating_point#Representable_numbers.2C_conversion_and_rounding" target="_top">representable</a>)
can be achieved by <code class="computeroutput"><span class="keyword">static_cast</span></code>ing
from a higher precision type, typically a <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
type like <code class="computeroutput"><span class="identifier">cpp_bin_float_50</span></code>,
but at the cost of increasing run-time 100-fold; this has been used here to
provide some of our reference values for testing.
</p>
<p>
For example, we get a reference value using a high precision type, for example;
</p>
<pre class="programlisting"><span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">multiprecision</span><span class="special">::</span><span class="identifier">cpp_bin_float_50</span><span class="special">;</span>
</pre>
<p>
that uses Halley iteration to refine until it is as precise as possible for
this <code class="computeroutput"><span class="identifier">cpp_bin_float_50</span></code> type.
</p>
<p>
As a further check we can compare this with a <a href="http://www.wolframalpha.com/" target="_top">Wolfram
Alpha</a> computation using command <code class="literal">N[ProductLog[10.], 50]</code>
to get 50 decimal digits and similarly <code class="literal">N[ProductLog[10.], 17]</code>
to get the nearest representable for 64-bit <code class="computeroutput"><span class="keyword">double</span></code>
precision.
</p>
<pre class="programlisting"> <span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">multiprecision</span><span class="special">::</span><span class="identifier">cpp_bin_float_50</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">float_distance</span><span class="special">;</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">z</span><span class="special">(</span><span class="string">"10."</span><span class="special">);</span> <span class="comment">// Note use a decimal digit string, not a double 10.</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">r</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">cpp_bin_float_50</span><span class="special">>::</span><span class="identifier">digits10</span><span class="special">);</span>
<span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span> <span class="comment">// Default policy.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(z) cpp_bin_float_50 = "</span> <span class="special"><<</span> <span class="identifier">r</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">//lambert_w0(z) cpp_bin_float_50 = 1.7455280027406993830743012648753899115352881290809</span>
<span class="comment">// [N[productlog[10], 50]] == 1.7455280027406993830743012648753899115352881290809</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"lambert_w0(z) static_cast from cpp_bin_float_50 = "</span>
<span class="special"><<</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">r</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// double lambert_w0(z) static_cast from cpp_bin_float_50 = 1.7455280027406994</span>
<span class="comment">// [N[productlog[10], 17]] == 1.7455280027406994</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"bits different from Wolfram = "</span>
<span class="special"><<</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">int</span><span class="special">>(</span><span class="identifier">float_distance</span><span class="special">(</span><span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">r</span><span class="special">),</span> <span class="number">1.7455280027406994</span><span class="special">))</span>
<span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// 0</span>
</pre>
<p>
giving us the same nearest representable using 64-bit <code class="computeroutput"><span class="keyword">double</span></code>
as <code class="computeroutput"><span class="number">1.7455280027406994</span></code>.
</p>
<p>
However, the rational polynomial and Fukushima Schroder approximations are
so good for type <code class="computeroutput"><span class="keyword">float</span></code> and <code class="computeroutput"><span class="keyword">double</span></code> that negligible improvement is gained
from a <code class="computeroutput"><span class="keyword">double</span></code> Halley step.
</p>
<p>
This is shown with <a href="../../../example/lambert_w_precision_example.cpp" target="_top">lambert_w_precision_example.cpp</a>
for Lambert <span class="emphasis"><em>W</em></span><sub>0</sub>:
</p>
<pre class="programlisting"><span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">lambert_w_detail</span><span class="special">::</span><span class="identifier">lambert_w_halley_step</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">epsilon_difference</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">relative_difference</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">showpoint</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// and show any significant trailing zeros too.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 17 decimal digits for double.</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">z50</span><span class="special">(</span><span class="string">"1.23"</span><span class="special">);</span> <span class="comment">// Note: use a decimal digit string, not a double 1.23!</span>
<span class="keyword">double</span> <span class="identifier">z</span> <span class="special">=</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">z50</span><span class="special">);</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">w50</span><span class="special">;</span>
<span class="identifier">w50</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z50</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">cpp_bin_float_50</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 50 decimal digits.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Reference Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") =\n "</span>
<span class="special"><<</span> <span class="identifier">w50</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 17 decimal digits for double.</span>
<span class="keyword">double</span> <span class="identifier">wr</span> <span class="special">=</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">w50</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Reference Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">wr</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">w</span> <span class="special">=</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Rat/poly Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// Add a Halley step to the value obtained from rational polynomial approximation.</span>
<span class="keyword">double</span> <span class="identifier">ww</span> <span class="special">=</span> <span class="identifier">lambert_w_halley_step</span><span class="special">(</span><span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">),</span> <span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Halley Step Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">lambert_w_halley_step</span><span class="special">(</span><span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">),</span> <span class="identifier">z</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"absolute difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">w</span> <span class="special">-</span> <span class="identifier">ww</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon for float = "</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">epsilon</span><span class="special">()</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"bits different from Halley step = "</span> <span class="special"><<</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">int</span><span class="special">>(</span><span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">))</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
with this output:
</p>
<pre class="programlisting"><span class="identifier">Reference</span> <span class="identifier">Lambert</span> <span class="identifier">W</span> <span class="special">(</span><span class="number">1.2299999999999999822364316059974953532218933105468750</span><span class="special">)</span> <span class="special">=</span>
<span class="number">0.64520356959320237759035605255334853830173300262666480</span>
<span class="identifier">Reference</span> <span class="identifier">Lambert</span> <span class="identifier">W</span> <span class="special">(</span><span class="number">1.2300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="number">0.64520356959320235</span>
<span class="identifier">Rat</span><span class="special">/</span><span class="identifier">poly</span> <span class="identifier">Lambert</span> <span class="identifier">W</span> <span class="special">(</span><span class="number">1.2300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="number">0.64520356959320224</span>
<span class="identifier">Halley</span> <span class="identifier">Step</span> <span class="identifier">Lambert</span> <span class="identifier">W</span> <span class="special">(</span><span class="number">1.2300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="number">0.64520356959320235</span>
<span class="identifier">absolute</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="special">-</span><span class="number">1.1102230246251565e-16</span>
<span class="identifier">relative</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">1.7207329236029286e-16</span>
<span class="identifier">epsilon</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">0.77494921535422934</span>
<span class="identifier">epsilon</span> <span class="keyword">for</span> <span class="keyword">float</span> <span class="special">=</span> <span class="number">2.2204460492503131e-16</span>
<span class="identifier">bits</span> <span class="identifier">different</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">1</span>
</pre>
<p>
and then for <span class="emphasis"><em>W</em></span><sub>-1</sub>:
</p>
<pre class="programlisting"><span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">lambert_w_detail</span><span class="special">::</span><span class="identifier">lambert_w_halley_step</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">epsilon_difference</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">relative_difference</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">showpoint</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span> <span class="comment">// and show any significant trailing zeros too.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 17 decimal digits for double.</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">z50</span><span class="special">(</span><span class="string">"-0.123"</span><span class="special">);</span> <span class="comment">// Note: use a decimal digit string, not a double -1.234!</span>
<span class="keyword">double</span> <span class="identifier">z</span> <span class="special">=</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">z50</span><span class="special">);</span>
<span class="identifier">cpp_bin_float_50</span> <span class="identifier">wm1_50</span><span class="special">;</span>
<span class="identifier">wm1_50</span> <span class="special">=</span> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z50</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">cpp_bin_float_50</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 50 decimal digits.</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Reference Lambert W-1 ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") =\n "</span>
<span class="special"><<</span> <span class="identifier">wm1_50</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span> <span class="comment">// 17 decimal digits for double.</span>
<span class="keyword">double</span> <span class="identifier">wr</span> <span class="special">=</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">>(</span><span class="identifier">wm1_50</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Reference Lambert W-1 ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">wr</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="keyword">double</span> <span class="identifier">w</span> <span class="special">=</span> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Rat/poly Lambert W-1 ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="comment">// Add a Halley step to the value obtained from rational polynomial approximation.</span>
<span class="keyword">double</span> <span class="identifier">ww</span> <span class="special">=</span> <span class="identifier">lambert_w_halley_step</span><span class="special">(</span><span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z</span><span class="special">),</span> <span class="identifier">z</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"Halley Step Lambert W ("</span> <span class="special"><<</span> <span class="identifier">z</span> <span class="special"><<</span> <span class="string">") = "</span> <span class="special"><<</span> <span class="identifier">lambert_w_halley_step</span><span class="special">(</span><span class="identifier">lambert_wm1</span><span class="special">(</span><span class="identifier">z</span><span class="special">),</span> <span class="identifier">z</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"absolute difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">w</span> <span class="special">-</span> <span class="identifier">ww</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon difference from Halley step = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon for float = "</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">epsilon</span><span class="special">()</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"bits different from Halley step = "</span> <span class="special"><<</span> <span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">int</span><span class="special">>(</span><span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">w</span><span class="special">,</span> <span class="identifier">ww</span><span class="special">))</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
with this output:
</p>
<pre class="programlisting"><span class="identifier">Reference</span> <span class="identifier">Lambert</span> <span class="identifier">W</span><span class="special">-</span><span class="number">1</span> <span class="special">(-</span><span class="number">0.12299999999999999822364316059974953532218933105468750</span><span class="special">)</span> <span class="special">=</span>
<span class="special">-</span><span class="number">3.2849102557740360179084675531714935199110302996513384</span>
<span class="identifier">Reference</span> <span class="identifier">Lambert</span> <span class="identifier">W</span><span class="special">-</span><span class="number">1</span> <span class="special">(-</span><span class="number">0.12300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="special">-</span><span class="number">3.2849102557740362</span>
<span class="identifier">Rat</span><span class="special">/</span><span class="identifier">poly</span> <span class="identifier">Lambert</span> <span class="identifier">W</span><span class="special">-</span><span class="number">1</span> <span class="special">(-</span><span class="number">0.12300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="special">-</span><span class="number">3.2849102557740357</span>
<span class="identifier">Halley</span> <span class="identifier">Step</span> <span class="identifier">Lambert</span> <span class="identifier">W</span> <span class="special">(-</span><span class="number">0.12300000000000000</span><span class="special">)</span> <span class="special">=</span> <span class="special">-</span><span class="number">3.2849102557740362</span>
<span class="identifier">absolute</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">4.4408920985006262e-16</span>
<span class="identifier">relative</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">1.3519066740696092e-16</span>
<span class="identifier">epsilon</span> <span class="identifier">difference</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="number">0.60884463935795785</span>
<span class="identifier">epsilon</span> <span class="keyword">for</span> <span class="keyword">float</span> <span class="special">=</span> <span class="number">2.2204460492503131e-16</span>
<span class="identifier">bits</span> <span class="identifier">different</span> <span class="identifier">from</span> <span class="identifier">Halley</span> <span class="identifier">step</span> <span class="special">=</span> <span class="special">-</span><span class="number">1</span>
</pre>
<h6>
<a name="math_toolkit.lambert_w.h9"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.differences_distribution"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.differences_distribution">Distribution
of differences from 'best' <code class="computeroutput"><span class="keyword">double</span></code>
evaluations</a>
</h6>
<p>
The distribution of differences from 'best' are shown in these graphs comparing
<code class="computeroutput"><span class="keyword">double</span></code> precision evaluations with
reference 'best' z50 evaluations using <code class="computeroutput"><span class="identifier">cpp_bin_float_50</span></code>
type reduced to <code class="computeroutput"><span class="keyword">double</span></code> with <code class="computeroutput"><span class="keyword">static_cast</span><span class="special"><</span><span class="keyword">double</span><span class="special">(</span><span class="identifier">z50</span><span class="special">)</span></code> :
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_w0_errors_graph.svg" align="middle"></span>
</p></blockquote></div>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_wm1_errors_graph.svg" align="middle"></span>
</p></blockquote></div>
<p>
As noted in the implementation section, the distribution of these differences
is somewhat biased for Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> and this might be reduced
using a <code class="computeroutput"><span class="keyword">double</span></code> Halley step at
small runtime cost. But if you are seriously concerned to get really precise
computations, the only way is using a higher precision type and then reduce
to the desired type. Fortunately, <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
makes this very easy to program, if much slower.
</p>
<h5>
<a name="math_toolkit.lambert_w.h10"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.edge_cases"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.edge_cases">Edge
and Corner cases</a>
</h5>
<h6>
<a name="math_toolkit.lambert_w.h11"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.w0_edges"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.w0_edges">The
<span class="emphasis"><em>W</em></span><sub>0</sub> Branch</a>
</h6>
<p>
The domain of <span class="emphasis"><em>W</em></span><sub>0</sub> is [-<span class="emphasis"><em>e</em></span><sup>-1</sup>, ∞). Numerically,
</p>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_w0</span><span class="special">(-</span><span class="number">1</span><span class="special">/</span><span class="identifier">e</span><span class="special">)</span></code> is exactly -1.
</li>
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">z</span><span class="special">)</span></code> for
<code class="computeroutput"><span class="identifier">z</span> <span class="special"><</span>
<span class="special">-</span><span class="number">1</span><span class="special">/</span><span class="identifier">e</span></code> throws
a <code class="computeroutput"><span class="identifier">domain_error</span></code>, or returns
<code class="computeroutput"><span class="identifier">NaN</span></code> according to the policy.
</li>
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_w0</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">infinity</span><span class="special">())</span></code>
throws an <code class="computeroutput"><span class="identifier">overflow_error</span></code>.
</li>
</ul></div>
<p>
(An infinite argument probably indicates that something has already gone wrong,
but if it is desired to return infinity, this case should be handled before
calling <code class="computeroutput"><span class="identifier">lambert_w0</span></code>).
</p>
<h6>
<a name="math_toolkit.lambert_w.h12"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.wm1_edges"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.wm1_edges"><span class="emphasis"><em>W</em></span><sub>-1</sub> Branch</a>
</h6>
<p>
The domain of <span class="emphasis"><em>W</em></span><sub>-1</sub> is [-<span class="emphasis"><em>e</em></span><sup>-1</sup>, 0). Numerically,
</p>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_wm1</span><span class="special">(-</span><span class="number">1</span><span class="special">/</span><span class="identifier">e</span><span class="special">)</span></code> is exactly -1.
</li>
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_wm1</span><span class="special">(</span><span class="number">0</span><span class="special">)</span></code> returns
-∞ (or the nearest equivalent if <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">has_infinity</span>
<span class="special">==</span> <span class="keyword">false</span></code>).
</li>
<li class="listitem">
<code class="computeroutput"><span class="identifier">lambert_wm1</span><span class="special">(-</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">min</span><span class="special">())</span></code>
returns the maximum (most negative) possible value of Lambert <span class="emphasis"><em>W</em></span>
for the type T. <br> For example, for <code class="computeroutput"><span class="keyword">double</span></code>:
lambert_wm1(-2.2250738585072014e-308) = -714.96865723796634 <br> and
for <code class="computeroutput"><span class="keyword">float</span></code>: lambert_wm1(-1.17549435e-38)
= -91.8567734 <br>
</li>
<li class="listitem">
<p class="simpara">
<code class="computeroutput"><span class="identifier">z</span> <span class="special"><</span>
<span class="special">-</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">min</span><span class="special">()</span></code>, means that z is zero or denormalized
(if <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">has_denorm_min</span> <span class="special">==</span>
<span class="keyword">true</span></code>), for example: <code class="computeroutput"><span class="identifier">r</span> <span class="special">=</span> <span class="identifier">lambert_wm1</span><span class="special">(-</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">denorm_min</span><span class="special">());</span></code>
and an overflow_error exception is thrown, and will give a message like:
</p>
<p class="simpara">
Error in function boost::math::lambert_wm1<RealType>(<RealType>):
Argument z = -4.9406564584124654e-324 is too small (z < -std::numeric_limits<T>::min
so denormalized) for Lambert W-1 branch!
</p>
</li>
</ul></div>
<p>
Denormalized values are not supported for Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> (because
not all floating-point types denormalize), and anyway it only covers a tiny
fraction of the range of possible z arguments values.
</p>
<h5>
<a name="math_toolkit.lambert_w.h13"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.compilers"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.compilers">Compilers</a>
</h5>
<p>
The <code class="computeroutput"><span class="identifier">lambert_w</span><span class="special">.</span><span class="identifier">hpp</span></code> code has been shown to work on most C++98
compilers. (Apart from requiring C++11 extensions for using of <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><>::</span><span class="identifier">max_digits10</span></code>
in some diagnostics. Many old pre-c++11 compilers provide this extension but
may require enabling to use, for example using b2/bjam the lambert_w examples
use this command:
</p>
<pre class="programlisting"><span class="special">[</span> <span class="identifier">run</span> <span class="identifier">lambert_w_basic_example</span><span class="special">.</span><span class="identifier">cpp</span> <span class="special">:</span> <span class="special">:</span> <span class="special">:</span> <span class="special">[</span> <span class="identifier">requires</span> <span class="identifier">cxx11_numeric_limits</span> <span class="special">]</span> <span class="special">]</span>
</pre>
<p>
See <a href="../../../example/Jamfile.v2" target="_top">jamfile.v2</a>.)
</p>
<p>
For details of which compilers are expected to work see lambert_w tests and
examples in:<br> <a href="https://www.boost.org/development/tests/master/developer/math.html" target="_top">Boost
Test Summary report for master branch (used for latest release)</a><br>
<a href="https://www.boost.org/development/tests/develop/developer/math.html" target="_top">Boost
Test Summary report for latest developer branch</a>.
</p>
<p>
As expected, debug mode is very much slower than release.
</p>
<h6>
<a name="math_toolkit.lambert_w.h14"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.diagnostics"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.diagnostics">Diagnostics
Macros</a>
</h6>
<p>
Several macros are provided to output diagnostic information (potentially
<span class="bold"><strong>much</strong></span> output). These can be statements, for
example:
</p>
<p>
<code class="computeroutput"><span class="preprocessor">#define</span> <span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_TERMS</span></code>
</p>
<p>
placed <span class="bold"><strong>before</strong></span> the <code class="computeroutput"><span class="identifier">lambert_w</span></code>
include statement
</p>
<p>
<code class="computeroutput"><span class="preprocessor">#include</span> <span class="special"><</span><span class="identifier">boost</span><span class="special">/</span><span class="identifier">math</span><span class="special">/</span><span class="identifier">special_functions</span><span class="special">/</span><span class="identifier">lambert_w</span><span class="special">.</span><span class="identifier">hpp</span><span class="special">></span></code>,
</p>
<p>
or defined on the project compile command-line: <code class="computeroutput"><span class="special">/</span><span class="identifier">DBOOST_MATH_INSTRUMENT_LAMBERT_W_TERMS</span></code>,
</p>
<p>
or defined in a jamfile.v2: <code class="computeroutput"><span class="special"><</span><span class="identifier">define</span><span class="special">></span><span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_TERMS</span></code>
</p>
<pre class="programlisting"><span class="comment">// #define-able macros</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_HALLEY</span> <span class="comment">// Halley refinement diagnostics.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_PRECISION</span> <span class="comment">// Precision.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_WM1</span> <span class="comment">// W1 branch diagnostics.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_WM1_HALLEY</span> <span class="comment">// Halley refinement diagnostics only for W-1 branch.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_WM1_TINY</span> <span class="comment">// K > 64, z > -1.0264389699511303e-26</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_WM1_LOOKUP</span> <span class="comment">// Show results from W-1 lookup table.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_SCHROEDER</span> <span class="comment">// Schroeder refinement diagnostics.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_TERMS</span> <span class="comment">// Number of terms used for near-singularity series.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_SINGULARITY_SERIES</span> <span class="comment">// Show evaluation of series near branch singularity.</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_SMALL_Z_SERIES</span>
<span class="identifier">BOOST_MATH_INSTRUMENT_LAMBERT_W_SMALL_Z_SERIES_ITERATIONS</span> <span class="comment">// Show evaluation of series for small z.</span>
</pre>
<h5>
<a name="math_toolkit.lambert_w.h15"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.implementation"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.implementation">Implementation</a>
</h5>
<p>
There are many previous implementations, each with increasing accuracy and/or
speed. See <a class="link" href="lambert_w.html#math_toolkit.lambert_w.references">references</a>
below.
</p>
<p>
For most of the range of <span class="emphasis"><em>z</em></span> arguments, some initial approximation
followed by a single refinement, often using Halley or similar method, gives
a useful precision. For speed, several implementations avoid evaluation of
a iteration test using the exponential function, estimating that a single refinement
step will suffice, but these rarely get to the best result possible. To get
a better precision, additional refinements, probably iterative, are needed
for example, using <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.halley">Halley</a>
or <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.schroder">Schröder</a> methods.
</p>
<p>
For C++, the most precise results possible, closest to the nearest <a href="http://en.wikipedia.org/wiki/Floating_point#Representable_numbers.2C_conversion_and_rounding" target="_top">representable</a>
for the C++ type being used, it is usually necessary to use a higher precision
type for intermediate computation, finally static-casting back to the smaller
desired result type. This strategy is used by <a href="https://www.maplesoft.com" target="_top">Maple</a>
and <a href="http://www.wolframalpha.com/" target="_top">Wolfram Alpha</a>, for example,
using arbitrary precision arithmetic, and some of their high-precision values
are used for testing this library. This method is also used to provide some
<a href="https://www.boost.org/doc/libs/release/libs/test/doc/html/index.html" target="_top">Boost.Test</a>
values using <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>,
typically, a 50 decimal digit type like <code class="computeroutput"><span class="identifier">cpp_bin_float_50</span></code>
<code class="computeroutput"><span class="keyword">static_cast</span></code> to a <code class="computeroutput"><span class="keyword">float</span></code>, <code class="computeroutput"><span class="keyword">double</span></code>
or <code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
type.
</p>
<p>
For <span class="emphasis"><em>z</em></span> argument values near the singularity and near zero,
other approximations may be used, possibly followed by refinement or increasing
number of series terms until a desired precision is achieved. At extreme arguments
near to zero or the singularity at the branch point, even this fails and the
only method to achieve a really close result is to cast from a higher precision
type.
</p>
<p>
In practical applications, the increased computation required (often towards
a thousand-fold slower and requiring much additional code for <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>)
is not justified and the algorithms here do not implement this. But because
the Boost.Lambert_W algorithms has been tested using <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>,
users who require this can always easily achieve the nearest representation
for <a href="http://en.cppreference.com/w/cpp/language/types" target="_top">fundamental
(built-in) types</a> - if the application justifies the very large extra
computation cost.
</p>
<h6>
<a name="math_toolkit.lambert_w.h16"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.evolution_of_this_implementation"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.evolution_of_this_implementation">Evolution
of this implementation</a>
</h6>
<p>
One compact real-only implementation was based on an algorithm by <a href="http://discovery.ucl.ac.uk/1482128/1/Luu_thesis.pdf" target="_top">Thomas
Luu, Thesis, University College London (2015)</a>, (see routine 11 on page
98 for his Lambert W algorithm) and his Halley refinement is used iteratively
when required. A first implementation was based on Thomas Luu's code posted
at <a href="https://svn.boost.org/trac/boost/ticket/11027" target="_top">Boost Trac #11027</a>.
It has been implemented from Luu's algorithm but templated on <code class="computeroutput"><span class="identifier">RealType</span></code> parameter and result and handles
both <a href="http://en.cppreference.com/w/cpp/language/types" target="_top">fundamental
(built-in) types</a> (<code class="computeroutput"><span class="keyword">float</span><span class="special">,</span> <span class="keyword">double</span><span class="special">,</span>
<span class="keyword">long</span> <span class="keyword">double</span></code>),
<a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>,
and also has been tested successfully with a proposed fixed_point type.
</p>
<p>
A first approximation was computed using the method of Barry et al (see references
5 & 6 below). This was extended to the widely used <a href="https://people.sc.fsu.edu/~jburkardt/f_src/toms443/toms443.html" target="_top">TOMS443</a>
FORTRAN and C++ versions by John Burkardt using Schroeder refinement(s). (For
users only requiring an accuracy of relative accuracy of 0.02%, Barry's function
alone might suffice, but a better <a href="https://en.wikipedia.org/wiki/Rational_function" target="_top">rational
function</a> approximation method has since been developed for this implementation).
</p>
<p>
We also considered using <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.newton">Newton-Raphson
iteration</a> method.
</p>
<pre class="programlisting"><span class="identifier">f</span><span class="special">(</span><span class="identifier">w</span><span class="special">)</span> <span class="special">=</span> <span class="identifier">w</span> <span class="identifier">e</span><span class="special">^</span><span class="identifier">w</span> <span class="special">-</span><span class="identifier">z</span> <span class="special">=</span> <span class="number">0</span> <span class="comment">// Luu equation 6.37</span>
<span class="identifier">f</span><span class="char">'(w) = e^w (1 + w), Wolfram alpha (d)/(dw)(f(w) = w exp(w) - z) = e^w (w + 1)
if (f(w) / f'</span><span class="special">(</span><span class="identifier">w</span><span class="special">)</span> <span class="special">-</span><span class="number">1</span> <span class="special"><</span> <span class="identifier">tolerance</span>
<span class="identifier">w1</span> <span class="special">=</span> <span class="identifier">w0</span> <span class="special">-</span> <span class="special">(</span><span class="identifier">expw0</span> <span class="special">*</span> <span class="special">(</span><span class="identifier">w0</span> <span class="special">+</span> <span class="number">1</span><span class="special">));</span> <span class="comment">// Refine new Newton/Raphson estimate.</span>
</pre>
<p>
but concluded that since the Newton-Raphson method takes typically 6 iterations
to converge within tolerance, whereas Halley usually takes only 1 to 3 iterations
to achieve an result within 1 <a href="http://en.wikipedia.org/wiki/Unit_in_the_last_place" target="_top">Unit
in the last place (ULP)</a>, so the Newton-Raphson method is unlikely to
be quicker than the additional cost of computing the 2nd derivative for Halley's
method.
</p>
<p>
Halley refinement uses the simplified formulae obtained from <a href="http://www.wolframalpha.com/input/?i=%5B2(z+exp(z)-w)+d%2Fdx+(z+exp(z)-w)%5D+%2F+%5B2+(d%2Fdx+(z+exp(z)-w))%5E2+-+(z+exp(z)-w)+d%5E2%2Fdx%5E2+(z+exp(z)-w)%5D" target="_top">Wolfram
Alpha</a>
</p>
<pre class="programlisting"><span class="special">[</span><span class="number">2</span><span class="special">(</span><span class="identifier">z</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">z</span><span class="special">)-</span><span class="identifier">w</span><span class="special">)</span> <span class="identifier">d</span><span class="special">/</span><span class="identifier">dx</span> <span class="special">(</span><span class="identifier">z</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">z</span><span class="special">)-</span><span class="identifier">w</span><span class="special">)]</span> <span class="special">/</span> <span class="special">[</span><span class="number">2</span> <span class="special">(</span><span class="identifier">d</span><span class="special">/</span><span class="identifier">dx</span> <span class="special">(</span><span class="identifier">z</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">z</span><span class="special">)-</span><span class="identifier">w</span><span class="special">))^</span><span class="number">2</span> <span class="special">-</span> <span class="special">(</span><span class="identifier">z</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">z</span><span class="special">)-</span><span class="identifier">w</span><span class="special">)</span> <span class="identifier">d</span><span class="special">^</span><span class="number">2</span><span class="special">/</span><span class="identifier">dx</span><span class="special">^</span><span class="number">2</span> <span class="special">(</span><span class="identifier">z</span> <span class="identifier">exp</span><span class="special">(</span><span class="identifier">z</span><span class="special">)-</span><span class="identifier">w</span><span class="special">)]</span>
</pre>
<h5>
<a name="math_toolkit.lambert_w.h17"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.compact_implementation"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.compact_implementation">Implementing
Compact Algorithms</a>
</h5>
<p>
The most compact algorithm can probably be implemented using the log approximation
of Corless et al. followed by Halley iteration (but is also slowest and least
precise near zero and near the branch singularity).
</p>
<h5>
<a name="math_toolkit.lambert_w.h18"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.faster_implementation"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.faster_implementation">Implementing
Faster Algorithms</a>
</h5>
<p>
More recently, the Tosio Fukushima has developed an even faster algorithm,
avoiding any transcendental function calls as these are necessarily expensive.
The current implementation of Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> is based on his
algorithm starting with a translation from Fukushima's FORTRAN into C++ by
Darko Veberic.
</p>
<p>
Many applications of the Lambert W function make many repeated evaluations
for Monte Carlo methods; for these applications speed is very important. Luu,
and Chapeau-Blondeau and Monir provide typical usage examples.
</p>
<p>
Fukushima improves the important observation that much of the execution time
of all previous iterative algorithms was spent evaluating transcendental functions,
usually <code class="computeroutput"><span class="identifier">exp</span></code>. He has put a lot
of work into avoiding any slow transcendental functions by using lookup tables
and bisection, finishing with a single Schroeder refinement, without any check
on the final precision of the result (necessarily evaluating an expensive exponential).
</p>
<p>
Theoretical and practical tests confirm that Fukushima's algorithm gives Lambert
W estimates with a known small error bound (several <a href="http://en.wikipedia.org/wiki/Unit_in_the_last_place" target="_top">Unit
in the last place (ULP)</a>) over nearly all the range of <span class="emphasis"><em>z</em></span>
argument.
</p>
<p>
A mean difference was computed to express the typical error and is often about
0.5 epsilon, the theoretical minimum. Using the <a href="../../../../../libs/math/doc/html/math_toolkit/next_float/float_distance.html" target="_top">Boost.Math
float_distance</a>, we can also express this as the number of bits that
are different from the nearest representable or 'exact' or 'best' value. The
number and distribution of these few bits differences was studied by binning,
including their sign. Bins for (signed) 0, 1, 2 and 3 and 4 bits proved suitable.
</p>
<p>
However, though these give results within a few <a href="http://en.wikipedia.org/wiki/Machine_epsilon" target="_top">machine
epsilon</a> of the nearest representable result, they do not get as close
as is very often possible with further refinement, nearly always to within
one or two <a href="http://en.wikipedia.org/wiki/Machine_epsilon" target="_top">machine
epsilon</a>.
</p>
<p>
More significantly, the evaluations of the sum of all signed differences using
the Fukshima algorithm show a slight bias, being more likely to be a bit or
few below the nearest representation than above; bias might have unwanted effects
on some statistical computations.
</p>
<p>
Fukushima's method also does not cover the full range of z arguments of 'float'
precision and above.
</p>
<p>
For this implementation of Lambert <span class="emphasis"><em>W</em></span><sub>0</sub>, John Maddock used
the Boost.Math <a href="http://en.wikipedia.org/wiki/Remez_algorithm" target="_top">Remez
algorithm</a> method program to devise a <a href="https://en.wikipedia.org/wiki/Rational_function" target="_top">rational
function</a> for several ranges of argument for the <span class="emphasis"><em>W</em></span><sub>0</sub> branch
of Lambert <span class="emphasis"><em>W</em></span> function. These minimax rational approximations
are combined for an algorithm that is both smaller and faster.
</p>
<p>
Sadly it has not proved practical to use the same <a href="http://en.wikipedia.org/wiki/Remez_algorithm" target="_top">Remez
algorithm</a> method for Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> branch and so
the Fukushima algorithm is retained for this branch.
</p>
<p>
An advantage of both minimax rational <a href="http://en.wikipedia.org/wiki/Remez_algorithm" target="_top">Remez
algorithm</a> approximations is that the <span class="bold"><strong>distribution</strong></span>
from the reference values is reasonably random and insignificantly biased.
</p>
<p>
For example, table below a test of Lambert <span class="emphasis"><em>W</em></span><sub>0</sub> 10000 values
of argument covering the main range of possible values, 10000 comparisons from
z = 0.0501 to 703, in 0.001 step factor 1.05 when module 7 == 0
</p>
<div class="table">
<a name="math_toolkit.lambert_w.lambert_w0_Fukushima"></a><p class="title"><b>Table 8.73. Fukushima Lambert <span class="emphasis"><em>W</em></span><sub>0</sub> and typical improvement from
a single Halley step.</b></p>
<div class="table-contents"><table class="table" summary="Fukushima Lambert W0 and typical improvement from
a single Halley step.">
<colgroup>
<col>
<col>
<col>
<col>
<col>
<col>
<col>
</colgroup>
<thead><tr>
<th>
<p>
Method
</p>
</th>
<th>
<p>
Exact
</p>
</th>
<th>
<p>
One_bit
</p>
</th>
<th>
<p>
Two_bits
</p>
</th>
<th>
<p>
Few_bits
</p>
</th>
<th>
<p>
inexact
</p>
</th>
<th>
<p>
bias
</p>
</th>
</tr></thead>
<tbody>
<tr>
<td>
<p>
Schroeder <span class="emphasis"><em>W</em></span><sub>0</sub>
</p>
</td>
<td>
<p>
8804
</p>
</td>
<td>
<p>
1154
</p>
</td>
<td>
<p>
37
</p>
</td>
<td>
<p>
5
</p>
</td>
<td>
<p>
1243
</p>
</td>
<td>
<p>
-1193
</p>
</td>
</tr>
<tr>
<td>
<p>
after Halley step
</p>
</td>
<td>
<p>
9710
</p>
</td>
<td>
<p>
288
</p>
</td>
<td>
<p>
2
</p>
</td>
<td>
<p>
0
</p>
</td>
<td>
<p>
292
</p>
</td>
<td>
<p>
22
</p>
</td>
</tr>
</tbody>
</table></div>
</div>
<br class="table-break"><p>
Lambert <span class="emphasis"><em>W</em></span><sub>0</sub> values computed using the Fukushima method with
Schroeder refinement gave about 1/6 <code class="computeroutput"><span class="identifier">lambert_w0</span></code>
values that are one bit different from the 'best', and < 1% that are a few
bits 'wrong'. If a Halley refinement step is added, only 1 in 30 are even one
bit different, and only 2 two-bits 'wrong'.
</p>
<div class="table">
<a name="math_toolkit.lambert_w.lambert_w0_plus_halley"></a><p class="title"><b>Table 8.74. Rational polynomial Lambert <span class="emphasis"><em>W</em></span><sub>0</sub> and typical improvement
from a single Halley step.</b></p>
<div class="table-contents"><table class="table" summary="Rational polynomial Lambert W0 and typical improvement
from a single Halley step.">
<colgroup>
<col>
<col>
<col>
<col>
<col>
<col>
<col>
</colgroup>
<thead><tr>
<th>
<p>
Method
</p>
</th>
<th>
<p>
Exact
</p>
</th>
<th>
<p>
One_bit
</p>
</th>
<th>
<p>
Two_bits
</p>
</th>
<th>
<p>
Few_bits
</p>
</th>
<th>
<p>
inexact
</p>
</th>
<th>
<p>
bias
</p>
</th>
</tr></thead>
<tbody>
<tr>
<td>
<p>
rational/polynomial
</p>
</td>
<td>
<p>
7135
</p>
</td>
<td>
<p>
2863
</p>
</td>
<td>
<p>
2
</p>
</td>
<td>
<p>
0
</p>
</td>
<td>
<p>
2867
</p>
</td>
<td>
<p>
-59
</p>
</td>
</tr>
<tr>
<td>
<p>
after Halley step
</p>
</td>
<td>
<p>
9724
</p>
</td>
<td>
<p>
273
</p>
</td>
<td>
<p>
3
</p>
</td>
<td>
<p>
0
</p>
</td>
<td>
<p>
279
</p>
</td>
<td>
<p>
5
</p>
</td>
</tr>
</tbody>
</table></div>
</div>
<br class="table-break"><p>
With the rational polynomial approximation method, there are a third one-bit
from the best and none more than two-bits. Adding a Halley step (or iteration)
reduces the number that are one-bit different from about a third down to one
in 30; this is unavoidable 'computational noise'. An extra Halley step would
double the runtime for a tiny gain and so is not chosen for this implementation,
but remains a option, as detailed above.
</p>
<p>
For the Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> branch, the Fukushima algorithm is
used.
</p>
<div class="table">
<a name="math_toolkit.lambert_w.lambert_wm1_fukushima"></a><p class="title"><b>Table 8.75. Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> using Fukushima algorithm.</b></p>
<div class="table-contents"><table class="table" summary="Lambert W-1 using Fukushima algorithm.">
<colgroup>
<col>
<col>
<col>
<col>
<col>
<col>
<col>
</colgroup>
<thead><tr>
<th>
<p>
Method
</p>
</th>
<th>
<p>
Exact
</p>
</th>
<th>
<p>
One_bit
</p>
</th>
<th>
<p>
Two_bits
</p>
</th>
<th>
<p>
Few_bits
</p>
</th>
<th>
<p>
inexact
</p>
</th>
<th>
<p>
bias
</p>
</th>
</tr></thead>
<tbody>
<tr>
<td>
<p>
Fukushima <span class="emphasis"><em>W</em></span><sub>-1</sub>
</p>
</td>
<td>
<p>
7167
</p>
</td>
<td>
<p>
2704
</p>
</td>
<td>
<p>
129
</p>
</td>
<td>
<p>
0
</p>
</td>
<td>
<p>
2962
</p>
</td>
<td>
<p>
-160
</p>
</td>
</tr>
<tr>
<td>
<p>
plus Halley step
</p>
</td>
<td>
<p>
7379
</p>
</td>
<td>
<p>
2529
</p>
</td>
<td>
<p>
92
</p>
</td>
<td>
<p>
0
</p>
</td>
<td>
<p>
2713
</p>
</td>
<td>
<p>
549
</p>
</td>
</tr>
</tbody>
</table></div>
</div>
<br class="table-break"><h6>
<a name="math_toolkit.lambert_w.h19"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.lookup_tables"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.lookup_tables">Lookup
tables</a>
</h6>
<p>
For speed during the bisection, Fukushima's algorithm computes lookup tables
of powers of e and z for integral Lambert W. There are 64 elements in these
tables. The FORTRAN version (and the C++ translation by Veberic) computed these
(once) as <code class="computeroutput"><span class="keyword">static</span></code> data. This is
slower, may cause trouble with multithreading, and is slightly inaccurate because
of rounding errors from repeated(64) multiplications.
</p>
<p>
In this implementation the array values have been computed using <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
50 decimal digit and output as C++ arrays 37 decimal digit <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code> literals using <code class="computeroutput"><span class="identifier">max_digits10</span></code> precision
</p>
<pre class="programlisting"><span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">cpp_bin_float_quad</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
</pre>
<p>
The arrays are as <code class="computeroutput"><span class="keyword">const</span></code> and <code class="computeroutput"><span class="keyword">constexpr</span></code> and <code class="computeroutput"><span class="keyword">static</span></code>
as possible (for the compiler version), using BOOST_STATIC_CONSTEXPR macro.
(See <a href="../../../tools/lambert_w_lookup_table_generator.cpp" target="_top">lambert_w_lookup_table_generator.cpp</a>
The precision was chosen to ensure that if used as <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code> arrays, then the values output
to <a href="../../../include/boost/math/special_functions/detail/lambert_w_lookup_table.ipp" target="_top">lambert_w_lookup_table.ipp</a>
will be the nearest representable value for the type chose by a <code class="computeroutput"><span class="keyword">typedef</span></code> in <a href="../../../include/boost/math/special_functions/lambert_w.hpp" target="_top">lambert_w.hpp</a>.
</p>
<pre class="programlisting"><span class="keyword">typedef</span> <span class="keyword">double</span> <span class="identifier">lookup_t</span><span class="special">;</span> <span class="comment">// Type for lookup table (`double` or `float`, or even `long double`?)</span>
</pre>
<p>
This is to allow for future use at higher precision, up to platforms that use
128-bit (hardware or software) for their <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code> type.
</p>
<p>
The accuracy of the tables was confirmed using <a href="http://www.wolframalpha.com/" target="_top">Wolfram
Alpha</a> and agrees at the 37th decimal place, so ensuring that the value
is exactly read into even 128-bit <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code> to the nearest representation.
</p>
<h6>
<a name="math_toolkit.lambert_w.h20"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.higher_precision"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.higher_precision">Higher
precision</a>
</h6>
<p>
For types more precise than <code class="computeroutput"><span class="keyword">double</span></code>,
Fukushima reported that it was best to use the <code class="computeroutput"><span class="keyword">double</span></code>
estimate as a starting point, followed by refinement using <a class="link" href="roots_deriv.html#math_toolkit.roots_deriv.halley">Halley</a>
iterations or other methods; our experience confirms this.
</p>
<p>
Using <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
it is simple to compute very high precision values of Lambert W at least to
thousands of decimal digits over most of the range of z arguments.
</p>
<p>
For this reason, the lookup tables and bisection are only carried out at low
precision, usually <code class="computeroutput"><span class="keyword">double</span></code>, chosen
by the <code class="computeroutput"><span class="keyword">typedef</span> <span class="keyword">double</span>
<span class="identifier">lookup_t</span></code>. Unlike the FORTRAN version,
the lookup tables of Lambert_W of integral values are precomputed as C++ static
arrays of floating-point literals. The default is a <code class="computeroutput"><span class="keyword">typedef</span></code>
setting the type to <code class="computeroutput"><span class="keyword">double</span></code>. To
allow users to vary the precision from <code class="computeroutput"><span class="keyword">float</span></code>
to <code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
these are computed to 128-bit precision to ensure that even platforms with
<code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
do not lose precision.
</p>
<p>
The FORTRAN version and translation only permits the z argument to be the largest
items in these lookup arrays, <code class="computeroutput"><span class="identifier">wm0s</span><span class="special">[</span><span class="number">64</span><span class="special">]</span>
<span class="special">=</span> <span class="number">3.99049</span></code>,
producing an error message and returning <code class="computeroutput"><span class="identifier">NaN</span></code>.
So 64 is the largest possible value ever returned from the <code class="computeroutput"><span class="identifier">lambert_w0</span></code>
function. This is far from the <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><>::</span><span class="identifier">max</span><span class="special">()</span></code> for even <code class="computeroutput"><span class="keyword">float</span></code>s.
Therefore this implementation uses an approximation or 'guess' and Halley's
method to refine the result. Logarithmic approximation is discussed at length
by R.M.Corless et al. (page 349). Here we use the first two terms of equation
4.19:
</p>
<pre class="programlisting"><span class="identifier">T</span> <span class="identifier">lz</span> <span class="special">=</span> <span class="identifier">log</span><span class="special">(</span><span class="identifier">z</span><span class="special">);</span>
<span class="identifier">T</span> <span class="identifier">llz</span> <span class="special">=</span> <span class="identifier">log</span><span class="special">(</span><span class="identifier">lz</span><span class="special">);</span>
<span class="identifier">guess</span> <span class="special">=</span> <span class="identifier">lz</span> <span class="special">-</span> <span class="identifier">llz</span> <span class="special">+</span> <span class="special">(</span><span class="identifier">llz</span> <span class="special">/</span> <span class="identifier">lz</span><span class="special">);</span>
</pre>
<p>
This gives a useful precision suitable for Halley refinement.
</p>
<p>
Similarly, for Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> branch, tiny values very near
zero, W > 64 cannot be computed using the lookup table. For this region,
an approximation followed by a few (usually 3) Halley refinements. See <a class="link" href="lambert_w.html#math_toolkit.lambert_w.wm1_near_zero">wm1_near_zero</a>.
</p>
<p>
For the less well-behaved regions for Lambert <span class="emphasis"><em>W</em></span><sub>0</sub> <span class="emphasis"><em>z</em></span>
arguments near zero, and near the branch singularity at <span class="emphasis"><em>-1/e</em></span>,
some series functions are used.
</p>
<h6>
<a name="math_toolkit.lambert_w.h21"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.small_z"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.small_z">Small
values of argument z near zero</a>
</h6>
<p>
When argument <span class="emphasis"><em>z</em></span> is small and near zero, there is an efficient
and accurate series evaluation method available (implemented in <code class="computeroutput"><span class="identifier">lambert_w0_small_z</span></code>). There is no equivalent
for the <span class="emphasis"><em>W</em></span><sub>-1</sub> branch as this only covers argument <code class="computeroutput"><span class="identifier">z</span> <span class="special"><</span> <span class="special">-</span><span class="number">1</span><span class="special">/</span><span class="identifier">e</span></code>.
The cutoff used <code class="computeroutput"><span class="identifier">abs</span><span class="special">(</span><span class="identifier">z</span><span class="special">)</span> <span class="special"><</span>
<span class="number">0.05</span></code> is as found by trial and error by
Fukushima.
</p>
<p>
Coefficients of the inverted series expansion of the Lambert W function around
<code class="computeroutput"><span class="identifier">z</span> <span class="special">=</span>
<span class="number">0</span></code> are computed following Fukushima using
17 terms of a Taylor series computed using <a href="http://www.wolfram.com/products/mathematica/index.html" target="_top">Wolfram
Mathematica</a> with
</p>
<pre class="programlisting"><span class="identifier">InverseSeries</span><span class="special">[</span><span class="identifier">Series</span><span class="special">[</span><span class="identifier">z</span> <span class="identifier">Exp</span><span class="special">[</span><span class="identifier">z</span><span class="special">],{</span><span class="identifier">z</span><span class="special">,</span><span class="number">0</span><span class="special">,</span><span class="number">17</span><span class="special">}]]</span>
</pre>
<p>
See Tosio Fukushima, Journal of Computational and Applied Mathematics 244 (2013),
page 86.
</p>
<p>
To provide higher precision constants (34 decimal digits) for types larger
than <code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>,
</p>
<pre class="programlisting"><span class="identifier">InverseSeries</span><span class="special">[</span><span class="identifier">Series</span><span class="special">[</span><span class="identifier">z</span> <span class="identifier">Exp</span><span class="special">[</span><span class="identifier">z</span><span class="special">],{</span><span class="identifier">z</span><span class="special">,</span><span class="number">0</span><span class="special">,</span><span class="number">34</span><span class="special">}]]</span>
</pre>
<p>
were also computed, but for current hardware it was found that evaluating a
<code class="computeroutput"><span class="keyword">double</span></code> precision and then refining
with Halley's method was quicker and more accurate.
</p>
<p>
Decimal values of specifications for built-in floating-point types below are
21 digits precision == <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">max_digits10</span></code> for <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code>.
</p>
<p>
Specializations for <code class="computeroutput"><span class="identifier">lambert_w0_small_z</span></code>
are provided for <code class="computeroutput"><span class="keyword">float</span></code>, <code class="computeroutput"><span class="keyword">double</span></code>, <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code>, <code class="computeroutput"><span class="identifier">float128</span></code>
and for <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
types.
</p>
<p>
The <code class="computeroutput"><span class="identifier">tag_type</span></code> selection is based
on the value <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">T</span><span class="special">>::</span><span class="identifier">max_digits10</span></code>
(and <span class="bold"><strong>not</strong></span> on the floating-point type T). This
distinguishes between <code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
types that commonly vary between 64 and 80-bits, and also compilers that have
a <code class="computeroutput"><span class="keyword">float</span></code> type using 64 bits and/or
<code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
using 128-bits.
</p>
<p>
As noted in the <a class="link" href="lambert_w.html#math_toolkit.lambert_w.implementation">implementation</a>
section above, it is only possible to ensure the nearest representable value
by casting from a higher precision type, computed at very, very much greater
cost.
</p>
<p>
For multiprecision types, first several terms of the series are tabulated and
evaluated as a polynomial: (this will save us a bunch of expensive calls to
<code class="computeroutput"><span class="identifier">pow</span></code>). Then our series functor
is initialized "as if" it had already reached term 18, enough evaluation
of built-in 64-bit double and float (and 80-bit <code class="computeroutput"><span class="keyword">long</span>
<span class="keyword">double</span></code>) types. Finally the functor is
called repeatedly to compute as many additional series terms as necessary to
achieve the desired precision, set from <code class="computeroutput"><span class="identifier">get_epsilon</span></code>
(or terminated by <code class="computeroutput"><span class="identifier">evaluation_error</span></code>
on reaching the set iteration limit <code class="computeroutput"><span class="identifier">max_series_iterations</span></code>).
</p>
<p>
A little more than one decimal digit of precision is gained by each additional
series term. This allows computation of Lambert W near zero to at least 1000
decimal digit precision, given sufficient compute time.
</p>
<h5>
<a name="math_toolkit.lambert_w.h22"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.near_singularity"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.near_singularity">Argument
z near the singularity at -1/e between branches <span class="emphasis"><em>W</em></span><sub>0</sub> and
<span class="emphasis"><em>W</em></span><sub>-1</sub> </a>
</h5>
<p>
Variants of Function <code class="computeroutput"><span class="identifier">lambert_w_singularity_series</span></code>
are used to handle <span class="emphasis"><em>z</em></span> arguments which are near to the singularity
at <code class="computeroutput"><span class="identifier">z</span> <span class="special">=</span>
<span class="special">-</span><span class="identifier">exp</span><span class="special">(-</span><span class="number">1</span><span class="special">)</span>
<span class="special">=</span> <span class="special">-</span><span class="number">3.6787944</span></code> where the branches <span class="emphasis"><em>W</em></span><sub>0</sub> and
<span class="emphasis"><em>W</em></span><sub>-1</sub> join.
</p>
<p>
T. Fukushima / Journal of Computational and Applied Mathematics 244 (2013)
77-89 describes using <a href="http://www.wolfram.com/products/mathematica/index.html" target="_top">Wolfram
Mathematica</a>
</p>
<pre class="programlisting"><span class="identifier">InverseSeries</span><span class="special">\[</span><span class="identifier">Series</span><span class="special">\[</span><span class="identifier">sqrt</span><span class="special">\[</span><span class="number">2</span><span class="special">(</span><span class="identifier">p</span> <span class="identifier">Exp</span><span class="special">\[</span><span class="number">1</span> <span class="special">+</span> <span class="identifier">p</span><span class="special">\]</span> <span class="special">+</span> <span class="number">1</span><span class="special">)\],</span> <span class="special">{</span><span class="identifier">p</span><span class="special">,-</span><span class="number">1</span><span class="special">,</span> <span class="number">20</span><span class="special">}\]\]</span>
</pre>
<p>
to provide his Table 3, page 85.
</p>
<p>
This implementation used <a href="http://www.wolfram.com/products/mathematica/index.html" target="_top">Wolfram
Mathematica</a> to obtain 40 series terms at 50 decimal digit precision
</p>
<pre class="programlisting"><span class="identifier">N</span><span class="special">\[</span><span class="identifier">InverseSeries</span><span class="special">\[</span><span class="identifier">Series</span><span class="special">\[</span><span class="identifier">Sqrt</span><span class="special">\[</span><span class="number">2</span><span class="special">(</span><span class="identifier">p</span> <span class="identifier">Exp</span><span class="special">\[</span><span class="number">1</span> <span class="special">+</span> <span class="identifier">p</span><span class="special">\]</span> <span class="special">+</span> <span class="number">1</span><span class="special">)\],</span> <span class="special">{</span> <span class="identifier">p</span><span class="special">,-</span><span class="number">1</span><span class="special">,</span><span class="number">40</span> <span class="special">}\]\],</span> <span class="number">50</span><span class="special">\]</span>
<span class="special">-</span><span class="number">1</span><span class="special">+</span><span class="identifier">p</span><span class="special">-</span><span class="identifier">p</span><span class="special">^</span><span class="number">2</span><span class="special">/</span><span class="number">3</span><span class="special">+(</span><span class="number">11</span> <span class="identifier">p</span><span class="special">^</span><span class="number">3</span><span class="special">)/</span><span class="number">72</span><span class="special">-(</span><span class="number">43</span> <span class="identifier">p</span><span class="special">^</span><span class="number">4</span><span class="special">)/</span><span class="number">540</span><span class="special">+(</span><span class="number">769</span> <span class="identifier">p</span><span class="special">^</span><span class="number">5</span><span class="special">)/</span><span class="number">17280</span><span class="special">-(</span><span class="number">221</span> <span class="identifier">p</span><span class="special">^</span><span class="number">6</span><span class="special">)/</span><span class="number">8505</span><span class="special">+(</span><span class="number">680863</span> <span class="identifier">p</span><span class="special">^</span><span class="number">7</span><span class="special">)/</span><span class="number">43545600</span> <span class="special">...</span>
</pre>
<p>
These constants are computed at compile time for the full precision for any
<code class="computeroutput"><span class="identifier">RealType</span> <span class="identifier">T</span></code>
using the original rationals from Fukushima Table 3.
</p>
<p>
Longer decimal digits strings are rationals pre-evaluated using <a href="http://www.wolfram.com/products/mathematica/index.html" target="_top">Wolfram
Mathematica</a>. Some integer constants overflow, so largest size available
is used, suffixed by <code class="computeroutput"><span class="identifier">uLL</span></code>.
</p>
<p>
Above the 14th term, the rationals exceed the range of <code class="computeroutput"><span class="keyword">unsigned</span>
<span class="keyword">long</span> <span class="keyword">long</span></code>
and are replaced by pre-computed decimal values at least 21 digits precision
== <code class="computeroutput"><span class="identifier">max_digits10</span></code> for <code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>.
</p>
<p>
A macro <code class="computeroutput"><span class="identifier">BOOST_MATH_TEST_VALUE</span></code>
(defined in <a href="../../../test/test_value.hpp" target="_top">test_value.hpp</a>)
taking a decimal floating-point literal was used to allow testing with both
built-in floating-point types like <code class="computeroutput"><span class="keyword">double</span></code>
which have constructors taking literal decimal values like <code class="computeroutput"><span class="number">3.14</span></code>,
<span class="bold"><strong>and</strong></span> also multiprecision and other User-defined
Types that only provide full-precision construction from decimal digit strings
like <code class="computeroutput"><span class="string">"3.14"</span></code>. (Construction
of multiprecision types from built-in floating-point types only provides the
precision of the built-in type, like <code class="computeroutput"><span class="keyword">double</span></code>,
only 17 decimal digits).
</p>
<div class="tip"><table border="0" summary="Tip">
<tr>
<td rowspan="2" align="center" valign="top" width="25"><img alt="[Tip]" src="../../../../../doc/src/images/tip.png"></td>
<th align="left">Tip</th>
</tr>
<tr><td align="left" valign="top"><p>
Be exceeding careful not to silently lose precision by constructing multiprecision
types from literal decimal types, usually <code class="literal">double</code>. Use
decimal digit strings like "3.1459" instead. See examples.
</p></td></tr>
</table></div>
<p>
Fukushima's implementation used 20 series terms; it was confirmed that using
more terms does not usefully increase accuracy.
</p>
<h6>
<a name="math_toolkit.lambert_w.h23"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.wm1_near_zero"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.wm1_near_zero">Lambert
<span class="emphasis"><em>W</em></span><sub>-1</sub> arguments values very near zero.</a>
</h6>
<p>
The lookup tables of Fukushima have only 64 elements, so that the z argument
nearest zero is -1.0264389699511303e-26, that corresponds to a maximum Lambert
<span class="emphasis"><em>W</em></span><sub>-1</sub> value of 64.0. Fukushima's implementation did not cater
for z argument values that are smaller (nearer to zero), but this implementation
adds code to accept smaller (but not denormalised) values of z. A crude approximation
for these very small values is to take the exponent and multiply by ln[10]
~= 2.3. We also tried the approximation first proposed by Corless et al. using
ln(-z), (equation 4.19 page 349) and then tried improving by a 2nd term -ln(ln(-z)),
and finally the ratio term -ln(ln(-z))/ln(-z).
</p>
<p>
For a z very close to z = -1.0264389699511303e-26 when W = 64, when effect
of ln(ln(-z) term, and ratio L1/L2 is greatest, the possible 'guesses' are
</p>
<pre class="programlisting"><span class="identifier">z</span> <span class="special">=</span> <span class="special">-</span><span class="number">1.e-26</span><span class="special">,</span> <span class="identifier">w</span> <span class="special">=</span> <span class="special">-</span><span class="number">64.02</span><span class="special">,</span> <span class="identifier">guess</span> <span class="special">=</span> <span class="special">-</span><span class="number">64.0277</span><span class="special">,</span> <span class="identifier">ln</span><span class="special">(-</span><span class="identifier">z</span><span class="special">)</span> <span class="special">=</span> <span class="special">-</span><span class="number">59.8672</span><span class="special">,</span> <span class="identifier">ln</span><span class="special">(-</span><span class="identifier">ln</span><span class="special">(-</span><span class="identifier">z</span><span class="special">)</span> <span class="special">=</span> <span class="number">4.0921</span><span class="special">,</span> <span class="identifier">llz</span><span class="special">/</span><span class="identifier">lz</span> <span class="special">=</span> <span class="special">-</span><span class="number">0.0684</span>
</pre>
<p>
whereas at the minimum (unnormalized) z
</p>
<pre class="programlisting"><span class="identifier">z</span> <span class="special">=</span> <span class="special">-</span><span class="number">2.2250e-308</span><span class="special">,</span> <span class="identifier">w</span> <span class="special">=</span> <span class="special">-</span><span class="number">714.9</span><span class="special">,</span> <span class="identifier">guess</span> <span class="special">=</span> <span class="special">-</span><span class="number">714.9687</span><span class="special">,</span> <span class="identifier">ln</span><span class="special">(-</span><span class="identifier">z</span><span class="special">)</span> <span class="special">=</span> <span class="special">-</span><span class="number">708.3964</span><span class="special">,</span> <span class="identifier">ln</span><span class="special">(-</span><span class="identifier">ln</span><span class="special">(-</span><span class="identifier">z</span><span class="special">)</span> <span class="special">=</span> <span class="number">6.5630</span><span class="special">,</span> <span class="identifier">llz</span><span class="special">/</span><span class="identifier">lz</span> <span class="special">=</span> <span class="special">-</span><span class="number">0.0092</span>
</pre>
<p>
Although the addition of the 3rd ratio term did not reduce the number of Halley
iterations needed, it might allow return of a better low precision estimate
<span class="bold"><strong>without any Halley iterations</strong></span>. For the worst
case near w = 64, the error in the 'guess' is 0.008, ratio 0.0001 or 1 in 10,000
digits 10 ~= 4. Two log evaluations are still needed, but is probably over
an order of magnitude faster.
</p>
<p>
Halley's method was then used to refine the estimate of Lambert <span class="emphasis"><em>W</em></span><sub>-1</sub> from
this guess. Experiments showed that although all approximations reached with
<a href="http://en.wikipedia.org/wiki/Unit_in_the_last_place" target="_top">Unit in the
last place (ULP)</a> of the closest representable value, the computational
cost of the log functions was easily paid by far fewer iterations (typically
from 8 down to 4 iterations for double or float).
</p>
<h6>
<a name="math_toolkit.lambert_w.h24"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.halley"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.halley">Halley
refinement</a>
</h6>
<p>
After obtaining a double approximation, for <code class="computeroutput"><span class="keyword">double</span></code>,
<code class="computeroutput"><span class="keyword">long</span> <span class="keyword">double</span></code>
and <code class="computeroutput"><span class="identifier">quad</span></code> 128-bit precision,
a single iteration should suffice because Halley iteration should triple the
precision with each step (as long as the function is well behaved - and it
is), and since we have at least half of the bits correct already, one Halley
step is ample to get to 128-bit precision.
</p>
<h6>
<a name="math_toolkit.lambert_w.h25"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.lambert_w_derivatives"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.lambert_w_derivatives">Lambert
W Derivatives</a>
</h6>
<p>
The derivatives are computed using the formulae in <a href="https://en.wikipedia.org/wiki/Lambert_W_function#Derivative" target="_top">Wikipedia</a>.
</p>
<h5>
<a name="math_toolkit.lambert_w.h26"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.testing"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.testing">Testing</a>
</h5>
<p>
Initial testing of the algorithm was done using a small number of spot tests.
</p>
<p>
After it was established that the underlying algorithm (including unlimited
Halley refinements with a tight terminating criterion) was correct, some tables
of Lambert W values were computed using a 100 decimal digit precision <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
<code class="computeroutput"><span class="identifier">cpp_dec_float_100</span></code> type and
saved as a C++ program that will initialise arrays of values of z arguments
and lambert_W0 (<code class="computeroutput"><span class="identifier">lambert_w_mp_high_values</span><span class="special">.</span><span class="identifier">ipp</span></code> and
<code class="computeroutput"><span class="identifier">lambert_w_mp_low_values</span><span class="special">.</span><span class="identifier">ipp</span></code> ).
</p>
<p>
(A few of these pairs were checked against values computed by Wolfram Alpha
to try to guard against mistakes; all those tested agreed to the penultimate
decimal place, so they can be considered reliable to at least 98 decimal digits
precision).
</p>
<p>
A macro <code class="computeroutput"><span class="identifier">BOOST_MATH_TEST_VALUE</span></code>
was used to allow tests with any real type, both <a href="http://en.cppreference.com/w/cpp/language/types" target="_top">fundamental
(built-in) types</a> and <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>.
(This is necessary because <a href="http://en.cppreference.com/w/cpp/language/types" target="_top">fundamental
(built-in) types</a> have a constructor from floating-point literals like
3.1459F, 3.1459 or 3.1459L whereas <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
types may lose precision unless constructed from decimal digits strings like
"3.1459").
</p>
<p>
The 100-decimal digits precision pairs were then used to assess the precision
of less-precise types, including <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
<code class="computeroutput"><span class="identifier">cpp_bin_float_quad</span></code> and <code class="computeroutput"><span class="identifier">cpp_bin_float_50</span></code>. <code class="computeroutput"><span class="keyword">static_cast</span></code>ing
from the high precision types should give the closest representable value of
the less-precise type; this is then be used to assess the precision of the
Lambert W algorithm.
</p>
<p>
Tests using confirm that over nearly all the range of z arguments, nearly all
estimates are the nearest <a href="http://en.wikipedia.org/wiki/Floating_point#Representable_numbers.2C_conversion_and_rounding" target="_top">representable</a>
value, a minority are within 1 <a href="http://en.wikipedia.org/wiki/Unit_in_the_last_place" target="_top">Unit
in the last place (ULP)</a> and only a very few 2 ULP.
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_w0_errors_graph.svg" align="middle"></span>
</p></blockquote></div>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="inlinemediaobject"><img src="../../graphs/lambert_wm1_errors_graph.svg" align="middle"></span>
</p></blockquote></div>
<p>
For the range of z arguments over the range -0.35 to 0.5, a different algorithm
is used, but the same technique of evaluating reference values using a <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
<code class="computeroutput"><span class="identifier">cpp_dec_float_100</span></code> was used.
For extremely small z arguments, near zero, and those extremely near the singularity
at the branch point, precision can be much lower, as might be expected.
</p>
<p>
See source at: <a href="../../../example/lambert_w_simple_examples.cpp" target="_top">lambert_w_simple_examples.cpp</a>
<a href="../../../test/test_lambert_w.cpp" target="_top">test_lambert_w.cpp</a> contains
routine tests using <a href="https://www.boost.org/doc/libs/release/libs/test/doc/html/index.html" target="_top">Boost.Test</a>.
<a href="../../../tools/lambert_w_errors_graph.cpp" target="_top">lambert_w_errors_graph.cpp</a>
generating error graphs.
</p>
<h6>
<a name="math_toolkit.lambert_w.h27"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.quadrature_testing"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.quadrature_testing">Testing
with quadrature</a>
</h6>
<p>
A further method of testing over a wide range of argument z values was devised
by Nick Thompson (cunningly also to test the recently written quadrature routines
including <a href="../../../../../libs/multiprecision/doc/html/index.html" target="_top">Boost.Multiprecision</a>
!). These are definite integral formulas involving the W function that are
exactly known constants, for example, LambertW0(1/(z²) == √(2π), see <a href="https://en.wikipedia.org/wiki/Lambert_W_function#Definite_integrals" target="_top">Definite
Integrals</a>. Some care was needed to avoid overflow and underflow as
the integral function must evaluate to a finite result over the entire range.
</p>
<h6>
<a name="math_toolkit.lambert_w.h28"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.other_implementations"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.other_implementations">Other
implementations</a>
</h6>
<p>
The Lambert W has also been discussed in a <a href="http://lists.boost.org/Archives/boost/2016/09/230819.php" target="_top">Boost
thread</a>.
</p>
<p>
This also gives link to a prototype version by which also gives complex results
<code class="literal">(x < -exp(-1)</code>, about -0.367879). <a href="https://github.com/CzB404/lambert_w/" target="_top">Balazs
Cziraki 2016</a> Physicist, PhD student at Eotvos Lorand University, ELTE
TTK Institute of Physics, Budapest. has also produced a prototype C++ library
that can compute the Lambert W function for floating point <span class="bold"><strong>and
complex number types</strong></span>. This is not implemented here but might be
completed in the future.
</p>
<h5>
<a name="math_toolkit.lambert_w.h29"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.acknowledgements"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.acknowledgements">Acknowledgements</a>
</h5>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
Thanks to Wolfram for use of their invaluable online Wolfram Alpha service.
</li>
<li class="listitem">
Thanks for Mark Chapman for performing offline Wolfram computations.
</li>
</ul></div>
<h5>
<a name="math_toolkit.lambert_w.h30"></a>
<span class="phrase"><a name="math_toolkit.lambert_w.references"></a></span><a class="link" href="lambert_w.html#math_toolkit.lambert_w.references">References</a>
</h5>
<div class="orderedlist"><ol class="orderedlist" type="1">
<li class="listitem">
NIST Digital Library of Mathematical Functions. <a href="http://dlmf.nist.gov/4.13.F1" target="_top">http://dlmf.nist.gov/4.13.F1</a>.
</li>
<li class="listitem">
<a href="http://www.orcca.on.ca/LambertW/" target="_top">Lambert W Poster</a>,
R. M. Corless, G. H. Gonnet, D. E. G. Hare, D. J. Jeffery and D. E. Knuth,
On the Lambert W function Advances in Computational Mathematics, Vol 5,
(1996) pp 329-359.
</li>
<li class="listitem">
<a href="https://people.sc.fsu.edu/~jburkardt/f_src/toms443/toms443.html" target="_top">TOMS443</a>,
Andrew Barry, S. J. Barry, Patricia Culligan-Hensley, Algorithm 743: WAPR
- A Fortran routine for calculating real values of the W-function,<br>
ACM Transactions on Mathematical Software, Volume 21, Number 2, June 1995,
pages 172-181.<br> BISECT approximates the W function using bisection
(GNU licence). Original FORTRAN77 version by Andrew Barry, S. J. Barry,
Patricia Culligan-Hensley, this version by C++ version by John Burkardt.
</li>
<li class="listitem">
<a href="https://people.sc.fsu.edu/~jburkardt/f_src/toms743/toms743.html" target="_top">TOMS743</a>
Fortran 90 (updated 2014).
</li>
</ol></div>
<p>
Initial guesses based on:
</p>
<div class="orderedlist"><ol class="orderedlist" type="1">
<li class="listitem">
R.M.Corless, G.H.Gonnet, D.E.G.Hare, D.J.Jeffrey, and D.E.Knuth, On the
Lambert W function, Adv.Comput.Math., vol. 5, pp. 329 to 359, (1996).
</li>
<li class="listitem">
D.A. Barry, J.-Y. Parlange, L. Li, H. Prommer, C.J. Cunningham, and F.
Stagnitti. Analytical approximations for real values of the Lambert W-function.
Mathematics and Computers in Simulation, 53(1), 95-103 (2000).
</li>
<li class="listitem">
D.A. Barry, J.-Y. Parlange, L. Li, H. Prommer, C.J. Cunningham, and F.
Stagnitti. Erratum to analytical approximations for real values of the
Lambert W-function. Mathematics and Computers in Simulation, 59(6):543-543,
2002.
</li>
<li class="listitem">
C++ <a href="https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#c-cplusplus-language-support" target="_top">CUDA
NVidia GPU C/C++ language support</a> version of Luu algorithm, <a href="https://github.com/thomasluu/plog/blob/master/plog.cu" target="_top">plog</a>.
</li>
<li class="listitem">
<a href="http://discovery.ucl.ac.uk/1482128/1/Luu_thesis.pdf" target="_top">Thomas
Luu, Thesis, University College London (2015)</a>, see routine 11,
page 98 for Lambert W algorithm.
</li>
<li class="listitem">
Having Fun with Lambert W(x) Function, Darko Veberic University of Nova
Gorica, Slovenia IK, Forschungszentrum Karlsruhe, Germany, J. Stefan Institute,
Ljubljana, Slovenia.
</li>
<li class="listitem">
François Chapeau-Blondeau and Abdelilah Monir, Numerical Evaluation of the
Lambert W Function and Application to Generation of Generalized Gaussian
Noise With Exponent 1/2, IEEE Transactions on Signal Processing, 50(9)
(2002) 2160 - 2165.
</li>
<li class="listitem">
Toshio Fukushima, Precise and fast computation of Lambert W-functions without
transcendental function evaluations, Journal of Computational and Applied
Mathematics, 244 (2013) 77-89.
</li>
<li class="listitem">
T.C. Banwell and A. Jayakumar, Electronic Letter, Feb 2000, 36(4), pages
291-2. Exact analytical solution for current flow through diode with series
resistance. <a href="https://doi.org/10.1049/el:20000301" target="_top">https://doi.org/10.1049/el:20000301</a>
</li>
<li class="listitem">
Princeton Companion to Applied Mathematics, 'The Lambert-W function', Section
1.3: Series and Generating Functions.
</li>
<li class="listitem">
Cleve Moler, Mathworks blog <a href="https://blogs.mathworks.com/cleve/2013/09/02/the-lambert-w-function/#bfba4e2d-e049-45a6-8285-fe8b51d69ce7" target="_top">The
Lambert W Function</a>
</li>
<li class="listitem">
Digital Library of Mathematical Function, <a href="https://dlmf.nist.gov/4.13" target="_top">Lambert
W function</a>.
</li>
</ol></div>
</div>
<table xmlns:rev="http://www.cs.rpi.edu/~gregod/boost/tools/doc/revision" width="100%"><tr>
<td align="left"></td>
<td align="right"><div class="copyright-footer">Copyright © 2006-2021 Nikhar Agrawal, Anton Bikineev, Matthew Borland,
Paul A. Bristow, Marco Guazzone, Christopher Kormanyos, Hubert Holin, Bruno
Lalande, John Maddock, Evan Miller, Jeremy Murphy, Matthew Pulver, Johan Råde,
Gautam Sewani, Benjamin Sobotta, Nicholas Thompson, Thijs van den Berg, Daryle
Walker and Xiaogang Zhang<p>
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at <a href="http://www.boost.org/LICENSE_1_0.txt" target="_top">http://www.boost.org/LICENSE_1_0.txt</a>)
</p>
</div></td>
</tr></table>
<hr>
<div class="spirit-nav">
<a accesskey="p" href="jacobi_theta/jacobi_theta4.html"><img src="../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../special.html"><img src="../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../index.html"><img src="../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="zetas.html"><img src="../../../../../doc/src/images/next.png" alt="Next"></a>
</div>
</body>
</html>